Docs

nokeep API docs

An OpenAI-compatible API for two uncensored models. Nothing is added to your prompt, nothing is cut from the reply, and only token counts are kept.

Checked 26 October 2026

Start here

Quickstart: your first call in a minute, with any OpenAI SDK.
The free test key: wire up your code or app before you buy anything.
Set up an app: SillyTavern, JanitorAI, Open WebUI, RisuAI, Agnai and KoboldAI Lite.

Words and pictures

Reads pictures as well as text, answers at once, and thinks when you ask it to. For stories, games, research and agents.

For less

Under half the price, for text. It thinks briefly before every answer.

Use DeepSeek for most work and whenever there’s a picture. Use GLM when price matters more than pictures. Compare them.

Nothing added. Your messages go to the model exactly as sent. There is no system prompt of ours. Fields a model can’t take are dropped and named in the x-nokeep-ignored header.
Nothing cut. Replies stream back as written. The one refusal is ours: a request that crosses a hard limit gets a 422 with a code, and costs nothing.
Nothing kept. One row per call, with counts only. See the row.

To	Read
Stream, think, read pictures, call tools	Streaming · Thinking · Pictures · Tool calls
Pay and stay in budget	Pricing and billing · Keys and caps · Rate limits
Look up a field or an error	Chat completions · Errors · Headers
Check it is up	Status, checked every hour